Legal claims defining the scope of protection, as filed with the USPTO.
1. An apparatus for compression of 3D graphics data and untyped data comprising: a graphical processing unit (GPU) including a data compression pipeline for typed 3D graphics data and untyped data, the data compression pipeline including: a data port coupled with one or more shader cores, wherein the data port is to convert typed 3D graphics data to a format for pixel data and is to allow transfer of untyped data without format conversion, and a 3D compression/decompression unit to provide for compression and decompression of both typed 3D data and untyped data, wherein untyped data is to be compressed to be stored to a memory subsystem and is to be decompressed to be read from the memory subsystem; wherein the apparatus is to utilize blocks based at least in part on data hashing to store 3D data and is to utilize fixed sequential blocks for storage of untyped data.
2. The apparatus of claim 1 , wherein for untyped data the apparatus is to convert a buffer for the untyped data to a stateful buffer, the stateful buffer to identify untyped data for compression.
3. The apparatus of claim 1 , wherein memory allocation of the untyped data is to be determined by software, and wherein a GPU driver is to determine whether the untyped data is to be compressed.
4. The apparatus of claim 3 , wherein the apparatus is to pass one or more hints regarding data compression of the untyped data to the GPU driver.
5. The apparatus of claim 4 , wherein the one or more hints include one or more of whether compression should be enabled for a buffer and a native data size that maps to the buffer.
6. The apparatus of claim 4 , wherein the GPU driver is to allocate an auxiliary buffer to store compression metadata for the compression of the untyped data.
7. The apparatus of claim 1 , wherein the data compression pipeline further includes a surface state cache to hold a surface state for untyped data.
8. The apparatus of claim 1 , wherein the apparatus is to provide decompression of compressed untyped data without copying the data.
9. The apparatus of claim 8 , wherein the apparatus is to change a description or pointer to a buffer for the untyped data from a compressed designation to an uncompressed designation.
10. The apparatus of claim 1 , wherein the untyped data is machine learning or deep learning data.
11. A non-transitory computer-readable storage medium having stored thereon data representing sequences of instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: receiving, at a data port coupled with one or more shader cores, untyped data for processing by a graphical processing unit (GPU), the GPU including a data compression pipeline for typed 3D graphics data and untyped data, the data port to convert typed 3D graphics data to a format for pixel data and allow transfer of untyped data without format conversion; determining by a GPU driver to compress the untyped data; and performing compression of untyped data by the data compression pipeline; wherein the data compression pipeline includes a 3D compression/decompression unit to provide for compression of both typed 3D graphics data and untyped data, wherein untyped data is compressed to be stored to a memory subsystem and decompressed to be read from the memory subsystem; and wherein the GPU is to utilize blocks based at least in part on data hashing to store 3D data and is to utilize fixed sequential blocks for storage of untyped data.
12. The medium of claim 11 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: receiving a request to read untyped data from system memory of the memory subsystem; reading a block of data from a memory surface of the system memory via a first surface state to one or more memory arrays; and writing back the block of data as uncompressed data to the memory surface via a second surface state.
13. The medium of claim 12 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: preventing any thread other than a current thread from accessing the block of data until processing is completed.
14. The medium of claim 12 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: upon the entire memory surface being written, writing to the memory surface to indicate the memory surface is not compressed.
15. The medium of claim 11 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: converting a buffer for the untyped data to a stateful buffer, the stateful buffer to identify untyped data for compression.
16. The medium of claim 11 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: passing one or more hints regarding data compression of the untyped data to the GPU driver.
17. The medium of claim 11 , further comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: allocating an auxiliary buffer to store compression metadata for the compression of the untyped data.
18. A processing system comprising: a central processing unit (CPU); a system memory; and a graphical processing unit (GPU) a graphical processing unit (GPU) including a data compression pipeline for typed 3D graphics data and untyped data, the data compression pipeline including: a data port coupled with one or more shader cores, wherein the data port is to convert typed 3D graphics data to a format for pixel data and is to allow transfer of untyped data without format conversion, and a 3D compression/decompression unit to provide for compression and decompression of both typed 3D data and untyped data, wherein untyped data is to be compressed to be stored to a memory subsystem and data is to be decompressed to be read from the memory subsystem; wherein the system is to utilize blocks based at least in part on data hashing to store 3D data and is to utilize fixed sequential blocks for storage of untyped data.
19. The system of claim 18 , wherein for untyped data the system is to convert a buffer for the untyped data to a stateful buffer, the stateful buffer to identify untyped data for compression.
20. The system of claim 18 , wherein the data compression pipeline further includes a surface state cache to hold a surface state for untyped data.
21. The system of claim 18 , wherein the system is to provide decompression of compressed untyped data without copying the data.
Unknown
January 28, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.